Last updated 2026-11-20
Acceptable use policy
18+ only
You must be at least 18 years old, and of legal age where you live, to use AbliteratedLLM. Don't use the service, or build products on it, for anyone younger.
Your responsibility
You are solely responsible for your prompts, the outputs you receive and how you use them, including complying with the laws that apply to you. Your prompts include any images and video you send to models that accept them, which the automated check below does not screen. We provide the models as-is and are not responsible for what users do with the service or with its outputs. The terms of service set out the full disclaimer and limitation of liability.
Apart from the automated check below, nobody reviews your prompts or outputs. We don't store them, except the sealed record described in the privacy policy.
No sexually explicit content
You must not use AbliteratedLLM to request or generate sexually explicit content: depictions of sexual acts, erotica, explicit sexual roleplay, or nudity presented for arousal. This applies on every model and every account, in any part of a request (including system prompts, character cards and images or video you send), and whatever the framing, including fiction, roleplay and hypotheticals.
Non-explicit content stays allowed: romance and relationships without explicit sexual detail, sex education, sexual and reproductive health, and medical, scientific, legal or news content about sex.
The automated check below does not screen for this content, so a request that breaches this rule can still go through. If we learn of a breach, for example from an abuse report, we may suspend or terminate the account as the terms of service allow. Sexual content involving minors is a separate category that the automated check below does cover.
What the service blocks automatically
Except for designated reseller accounts described below, an automated in-memory check scans the text of every prompt before it reaches the model, and the end of every output after it completes, on every model. It covers text only: images and video are not screened automatically. It refuses two kinds of content only: child sexual abuse material (CSAM) and clear, actionable instructions for mass-casualty weapons (CBRN weapons or mass-casualty explosives). If it can't give an answer on a prompt in time, the request fails with 503 safety_unavailable instead of going through unchecked. A blocked request returns 400 content_policy_violation. The output check runs after the response has been delivered, so it can't stop that response: it records child sexual abuse material it finds, and if it can't run, that output isn't checked.
Designated reseller accounts may be exempt from both automated checks when the reseller is responsible for screening its users under its agreement with us. The acceptable-use restrictions still apply. Requests on these accounts are not sent to our safety classifier and do not generate classifier-triggered incident records. This does not waive legally required preservation or reporting.
- Child sexual abuse material: sexual content involving minors, including fictional or text-only content and anyone appearing under 18, and grooming. The request is blocked and a sealed incident record is kept, which we report as the law requires (see the privacy policy). Because delivered outputs can't be recalled, a flagged output is recorded the same way.
- Mass-casualty weapons: clear, actionable instructions for chemical, biological, radiological or nuclear (CBRN) weapons, or explosives, capable of mass casualties. The request is refused and is not recorded.
Neither carries automatic action on your account.
Models labelled Stock model
These run their vendor's original weights. They are not abliterated, keep the vendor's own refusals and may decline some requests. The automated check above applies to them as to every other model; images and video sent to the stock models that accept them are not screened. See the privacy policy for how their data is handled.
Models labelled Official DeepSeek API
These are DeepSeek's stock models, served through DeepSeek's official API. They are not abliterated, keep DeepSeek's own refusals, and DeepSeek may decline some requests on its side. Prompts and outputs for these models are processed and stored by DeepSeek in the People's Republic of China; see the privacy policy.
If we got it wrong
Automated checks can make mistakes. If you believe a request was blocked in error, write to support@abliteratedllm.dev and include the x-request-id header from the blocked response.
Reporting abuse
To report abuse or illegal content, see the report page.
Never send us images or copies of child sexual abuse material. Describe where you found it and include any request id. Report the material itself to the NCMEC CyberTipline or to your local police.